| GPT-5.5 | Gemini 3 Ultra | |
|---|---|---|
| arenaElo | 1432 | 1441 |
| humanEval | 94.2 | 91 |
| sweBench | 66 | 66 |
| mmlu | 93 | 93.4 |
| gpqa | 68.7 | 72.6 |
| aime | 92 | 93 |
| Price in/out ($/M) | $12 / $48 | $18 / $72 |
Gemini 3 Ultra has the larger context (3M) and native video; GPT-5.5 has tighter coding and the richest tool-calling ecosystem. Both sit within 10 Elo points of each other. Pick by cloud vendor or by whether video matters to you.